A Contribution to Reliable Rail Transport: AI-Powered Real-Time Wheel Defect Detection

The railway system is a green mode of transport that is essential for contributing significantly to economic growth and development, enhancing accessibility, and easing regional integration. The safety and efficiency of railway transport systems depend on the condition of wheels, as deterioration of wheels is a major cause of both human life and financial loss. Consequently, real-time monitoring is essential for early detection and preventing failures. This study proposes a framework using You Only Look Once (YOLO) models and a transformer model for railway wheel defect detection. A high-resolution, well-annotated, and augmented dataset has been developed to address class imbalances and problems related to the lack of annotated data. A comprehensive analysis of multiple YOLO instance segmentation models and a transformer model, including YOLOv5, YOLOv7, YOLOv8, YOLOv9, YOLOv11 instance segmentation and real-time detection (RTD) transformer were trained using extensive hyperparameter tuning. Real-time detection transformer resulted in relatively superior accuracy. The model also demonstrated higher precision of 0.97, recall of 0.97, and mean average precision (mAP) of 0.98, alongside efficient real-time processing at 24 FPS and a latency under 50 milliseconds. In order to validate the practical applicability of the model, the optimized real-time detection transformer model was deployed on an edge device for real-time defect detection in operational railway environments. The deployed system demonstrated reliability and accuracy, offering a viable solution for enhancing railway safety and operational efficiency through advanced defect detection. In order to detect and prevent tragic accidents, this technology has the potential to be incorporated into the wider online railway system for efficient and real-time wheelset condition monitoring.